Searches performed in August 2011 identified 207 apps from the app stores for Android, Apple, Blackberry and Windows Phone, of which 204 were available for screening (Figure 2). We excluded 101 that either contained no health or asthma-related content (n = 57), targeted clinicians (n = 35), were not in English (n = 7), or could not be started (n = 2). Excluded apps are summarized in Additional file 2. Subsequent discussion is restricted to the 103 apps that met inclusion criteria. Most (n = 94) were designed for smartphones. Although all iPhone apps can run on the iPad tablet, we found eight that included specific customizations to take advantage of the large screen size and one that was a tablet-only app.
Summary of characteristics of included apps
Fifty-six apps were sources of written (n = 43) and multimedia (n = 13) information about asthma and its management [14–69]. The remaining 47 were tools supporting aspects of asthma self-management and included diaries (n = 29), assessment instruments (n = 17) and location-based alerts (n = 6) [70–115]. Although we neither expected nor required that assignment into these two main categories be mutually exclusive, we found no English-language apps that combined both information and management tools. (A typology is provided in Additional file 3.) Sixty-one percent (n = 34) of information apps and 96% (n = 45) of management tools focussed on conventional medical management of asthma. Nineteen apps considered exclusively CAM, while five integrated content addressing both approaches. Seven apps targeted either younger children [38, 89] or their parents [16, 27, 106–108]. None targeted adolescents or elderly patients.
After including any in-app purchases required to access asthma-specific content, the majority of apps (n = 76/103) were not free with a median cost of £1.49 (mean £1.85, range £0.61 to £8.99). Apps offering management tools were more likely to be free (n = 19/47) compared to those presenting health information (n = 8/56).
Apps presenting health information
Apps presenting health information are summarized in Additional file 4.
Comprehensiveness of asthma information
Excluding apps exclusively addressing CAM, 38 apps were evaluated for comprehensiveness of asthma information [14–22, 24–41, 43, 48, 49, 54–56, 58, 59, 63, 64, 67]. The basic nature of asthma, including the role of inflammation, symptoms and prognosis was the most commonly addressed (at least partially), by two-thirds of apps (n = 26, breakdown in Additional file 5). Allergen and trigger avoidance were discussed by 18 but only covered in depth by 2 [25, 63]. Less than two-fifths addressed recognition of exacerbations (n = 14), self-monitoring (n = 10) and inhaler techniques (n = 10). Customized aspects of asthma management, including the role of an action plan and the prioritization of treatment goals according to patient wishes, were addressed least frequently by seven and three apps, respectively. Three apps wholly addressed six of the eight domains and provided partial coverage of the remaining two [19, 59, 63].
Lay management of acute asthma
Of the 14 medical apps containing conventional information about recognition and management of acute asthma [20, 25, 28, 29, 33–37, 49, 58, 59, 63, 64], 7 provided specific guidance on lay management of an asthma attack. Although none addressed all aspects of the step-wise approach recommended by guidelines, six gave advice that was broadly consistent [33–36, 49, 63] but lacked specific instructions on the dose and frequency of reliever inhaler use (addressed by n = 2/6) or the appropriate medical service to contact (addressed by n = 2/6). One app provided guidance that substantially differed from recommendations [29].
Eight apps suggested CAM procedures for acute asthma management [35, 36, 48, 51, 52, 60, 62, 65]. None recommended using a beta-agonist reliever inhaler or seeking conventional medical help should an alternative emergency procedure prove ineffective, although two contained details of conventional emergency management in separate sections [35, 36].
Compliance of information with evidence-based recommendations
We identified 72 instances where apps addressed items from our pre-defined set of evidence-based recommendations. Of these app-statements, 40 were asserted in line with current guidance. In all other cases (n = 32), apps appeared to unequivocally recommend a particular course of action where there is current uncertainty. Statements concerning active and passive smoke avoidance (exacerbates symptoms, n = 12 and 13), weight reduction in obesity (beneficial for asthma symptoms, n = 7) and the potential utility of immunotherapy (can be considered where a specific allergen is identified, n = 2) were correctly asserted by all apps that mentioned them. Recommendations about behavioural strategies for the avoidance of air pollution (n = 9), fungal allergens (n = 9), removal (rather than control of) pets (n = 5) and cockroach control (n = 7) were all delivered more variably (breakdown in Additional file 6). An unequivocal recommendation for flu vaccination was made by five of six apps.
A small number of apps actively cautioned against allopathic medical management. Four apps [18, 52, 60, 66] recommended avoiding conventional medical management because of the risks of side effects, addiction and worsening of the condition.
Apps providing tools for the management of asthma
Diaries
Twenty-nine apps offered functions for patients to track their asthma (Additional file 7 [70–98]). Diaries differed in terms of the information that they captured and the options given to patients for manipulating the recorded data. While a small number of diaries captured either asthma symptoms (n = 2 [87, 88]) or peak flow (n = 2 [93, 94]) alone, the majority (n = 23) allowed both symptoms and peak flow values to be recorded as well as recent medication use (n = 24). Most apps relied on manual entry of data; however, one [92] was able to source values from a Bluetooth-enabled peak flow meter and another (available on both iPhone and Windows Phone) from a wireless inhaler [96, 97] (untested in this review). Fifty-nine percent of (n = 17 of 29) diary apps lacked data validation to prevent out-of-range values to be entered [71, 73–77, 79–82, 86, 90, 93–96, 98]. Five diaries allowed customized self-management plans [76, 77, 84, 89, 92] that included emergency care instructions and prescribing details for different classes of medication. Four [76, 77, 84, 92] used a three-step action plan with traffic light colouring consistent with guideline recommendations [11]. However, none were able to vary the number of steps in the action plan, nor the thresholds at which the action plan steps were triggered (50 and 80%). All four used peak flow values entered in the diary to trigger a display of steps to be taken by the patient based on their action plan. Although recommended by guidelines, none included an equivalent function based on recorded symptoms.
Five apps [82, 90, 95–97] provided a function to track the doses remaining in their pressured Metered-Dose Inhaler (pMDI). Each app used a similar approach, providing a visual warning when the device was running low.
Assessment instruments
The sources and scoring mechanisms of asthma status questionnaires embedded in seven [82, 89, 99, 100, 106–108] were reviewed (Additional file 8). Only one app cited the source [82], assigning a numeric score based on Global Initiative for Asthma criteria for asthma control [7]. However, while these criteria exist [7], we could find no validated approach that recommends assigning a numeric score to each criterion and presenting the result as an aggregate sum. One [99] used, without attribution, the adult and paediatric versions of a standard instrument, the Asthma Control Test [116, 117]. Scoring errors were found in the adult version, which meant that no matter how minimal an individual's current symptoms, the app would always recommend seeking medical help. We could not find validation information for any of the other tools.
Three iPhone apps [101–103] used the device microphone to analyse breath sounds and provided diagnostic commentary, for example, the identification of wheeze. We were unable to locate validation information for these diagnostic tools.
Seven apps incorporated predicted peak flow calculators as either a dedicated calculator (n = 3 [104, 105, 118]) or within a diary to generate reference values for charting (n = 4 [80, 81, 83, 94]). Only one of the calculators [118] provided attribution. We were able to identify the calculation algorithm for one other [105]. Both had bugs which resulted in incorrect output being generated under certain circumstances. One [105] would silently forget the gender of the patient and subsequently provide male predicted values if the device was physically rotated to change from a portrait to landscape screen display. The other had a systematic error where female predicted values were returned for individuals five inches shorter in height than those entered [118]. Despite writing to the publishers, we could not identify the underlying algorithm for the third calculator or any but one of the diary apps [83], the performance of which could not be verified because of problems entering data. Only one acknowledged the use of different peak flow measurement scales by allowing the user to pick which type of meter they used [94].
Other tools
Six apps provided location-based pollen or pollution alerts for users in the United States and Ireland [70, 109–113] (Additional file 9). One product - available as apps on both Blackberry [115] and Apple [114] devices - did not fit into the categories of tools described above, offering paid-for audio recordings of Indian chants intended for use by those with a range of conditions including asthma.
Compliance with health information best-practice principles
The purpose of the app was clearly stated or interpretable in 86% of health information apps (n = 48) and 96% of management tools (n = 45). Content authorship was stated in 18 of 56 (32%) health information apps. Six apps [19, 29, 35, 36, 48, 66] were eBook versions of texts originally available in hard copy. Where information was not attributed, we searched online in an attempt to locate any original source. A quarter of information apps (n = 14, of which 10 were paid for) used content available freely online without attribution, for example, from Wikipedia [20, 25]. In a further five cases [18, 31, 33, 34, 57], we found matched content online but it was unclear whether reproduction was authorized. The date of content creation was identified for only one app [27] and none provided a content expiry date. Only one provided details of its editorial policy through a linked website [27].
An explicit confidentiality policy - found either in the app or on an associated website - was identified for only 5 of 29 apps (17%) in which personal data could be recorded [70, 72, 82–84]. Four apps offered a password protection mechanism to assist in securing data [78, 83, 84, 96]. We were able to identify the funding source for the app in 23 cases: 2 were sponsored by local US government [109, 110]; one medication tracker [95] (and 2 German-language apps excluded from the analysis [119, 120]) were sponsored by pharmaceutical companies; 2 by a company developing an electronic inhaler [96, 97] and the rest by commercial companies Twenty-two apps incorporated advertisements but none detailed an advertising policy. Most (n = 17) were for products unrelated to health and the remainder promoted content offered by the same publisher. Third party endorsements were present for two apps; from the US National Institutes of Health [59] and the UK Department of Health [27], under the Information Quality Mark scheme. Fifty-five percent of all apps offered a means to contact the authors using either email (n = 41), an online form (n = 14) or an in-app form (n = 2).